CoreWeave Presents at Bank of America Global Technology Conference

On June 3, 2026, Nick Robbins, vice president of corporate development and investor relations at CoreWeave, Inc., presented at the Bank of America 2026 Global Technology Conference in San Francisco, with the session listed to start at 10:40 a.m. Pacific Time, per CoreWeave's investor-relations announcement. In the published session transcript, the moderator, Tal Liani of BofA Securities, framed the discussion around differentiation, customer value, longevity, and order visibility. Robbins highlighted customer diversification, noting that Microsoft, which represented roughly 85% of CoreWeave's revenue backlog at its IPO, is no longer the company's largest customer. CoreWeave's investor-relations page and press notices announced a live webcast and replay.

The transcript centered on market positioning rather than product specifics. Discussions at similar conferences typically concentrate on capacity growth, GPU supply and pricing dynamics, and how providers map hardware availability to customer workloads. For practitioners, those are the operational levers that affect utilization, throughput, and per-GPU economics.

The appearance occurred amid broad investor interest in GPU-backed cloud capacity; coverage noted Bank of America's favorable stance on GPU-focused data-center names, citing CoreWeave and Nebius among top picks. That framing reflects appetite for companies exposed to large-model training and inference demand. For infrastructure and ML finance teams, clarity from providers about order visibility and multi-quarter demand trends matters for capacity planning and total-cost-of-training estimates.

Observers will look for three indicators in upcoming disclosures: explicit capacity and utilization metrics (rack-level or GPU counts) in earnings or investor slides; commentary on customer mix and contract duration that affects revenue visibility; and GPU-pricing trends that drive gross margins. These are standard signals used to assess how supply-side constraints translate into available compute for model development.